A&A 601, A58 (2017)DOI: 10.1051/0004-6361/201730437c© ESO 2017

Astronomy&Astrophysics

Spectroscopic twin to the hypervelocity sdO star US 708 and threefast sdB stars from the Hyper-MUCHFUSS project

E. Ziegerer1, U. Heber1, S. Geier1, 2, 3, 4, A. Irrgang1, T. Kupfer5, F. Fürst5, 6, and J. Schaffenroth1

1 Dr. Karl Remeis-Observatory & ECAP, Astronomical Institute, Friedrich-Alexander University Erlangen-Nürnberg,Sternwartstr. 7, 96049 Bamberg, Germanye-mail: Eva.Ziegerer@fau.de

2 European Southern Observatory, Karl-Schwarzschild-Str. 2, 85748 Garching, Germany3 Department of Physics, University of Warwick, Coventry CV4 AL, UK4 Institute for Astronomy and Astrophysics, Kepler Center for Astro and Particle Physics, Eberhard Karls University, Sand 1,

72076 Tübingen, Germany5 Division of Physics, Mathematics, and Astronomy, California Institute of Technology, Passadena, CA 91125, USA6 European Space Astronomy Centre (ESA/ESAC), Operations Department, 28692 Villanueva de la Cañada (Madrid), Spain

Received 13 January 2017 / Accepted 26 March 2017

ABSTRACT

Important tracers for the dark matter halo of the Galaxy are hypervelocity stars (HVSs), which are faster than the local escapevelocity of the Galaxy and their slower counterparts, the high-velocity stars in the Galactic halo. Such HVSs are believed to beejected from the Galactic centre (GC) through tidal disruption of a binary by the super-massive black hole (Hills mechanism). TheHyper-MUCHFUSS survey aims at finding high-velocity potentially unbound hot subdwarf stars. We present the spectroscopic andkinematical analyses of a He-sdO as well as three candidates among the sdB stars using optical Keck/ESI and VLT (X-shooter, FORS)spectroscopy. Proper motions are determined by combining positions from early-epoch photographic plates with those derived frommodern digital sky surveys. The Galactic rest frame velocities range from 203 km s−1 to 660 km s−1, indicating that most likely all fourstars are gravitationally bound to the Galaxy. With Teff = 47 000 K and a surface gravity of log g = 5.7, SDSS J205030.39−061957.8(J2050) is a spectroscopic twin of the hypervelocity He-sdO US 708. As for the latter, the GC is excluded as a place of origin basedon the kinematic analysis. Hence, the Hills mechanism can be excluded for J2050. The ejection velocity is much more moderate(385 ± 79 km s−1) than that of US 708 (998 ± 68 km s−1). The binary thermonuclear supernova scenario suggested for US 708 wouldexplain the observed properties of J2050 very well without pushing the model parameters to their extreme limits, as required forUS 708. Accordingly, the star would be the surviving donor of a type Ia supernova. Three sdB stars also showed extreme kinematics;one could be a HVS ejected from the GC, whereas the other two could be ejected from the Galactic disk through the binary supernovamechanism. Alternatively, they might be extreme halo stars.

Key words. stars: kinematics and dynamics – subdwarfs – stars: atmospheres – Galaxy: halo

1. Introduction

Hypervelocity stars (HVSs) are stars that move so fast thatthey may exceed the escape velocity of the Galaxy. In the late1980s, it was predicted by Hills (1988) from numerical exper-iments that a star can be ejected from the Galaxy with veloc-ities exceeding the escape velocity by the disruption of a bi-nary through tidal interaction with a super-massive black hole(SMBH). The first such stars were discovered serendipitouslyin 2005 (Brown et al. 2005; Hirsch et al. 2005; Edelmann et al.2005). However, Brown et al. (2007) showed that about 50% ofthe ejected stars undergoing this mechanism remain bound tothe Galaxy. We use the term HVS only for stars that are trulyunbound. Interestingly, the nature, number, and distribution ofthe so-called S-stars, which are normal main-sequence B-starsin the central arcsecond of the Galaxy on close eccentric orbitsaround the SMBH, are consistent with expectations for the for-mer companions of HVS (Svensson et al. 2008; Madigan et al.2014).

In their survey for unbound stars, Brown et al. (2014) dis-covered 21 unbound HVSs and 17 lower velocity stars of spec-tral type B with masses between 2.5 and 4 M�, which means that

these stars have short lifetimes. The Galactic centre (GC) is theonly place in our Galaxy known to host an SMBH (Schödel et al.2003; Ghez et al. 2005; Gillessen et al. 2009), and therefore theGC is considered the likely place of origin of HVSs.

The Hills scenario has been studied in many variations. Thisincludes binary SMBHs, binaries consisting of an SMBH and anintermediate-mass black hole, triple star disruption, in-spiral ofa young stellar cluster forming jets of HVSs, and many other nu-merical calculations (for details we refer to the review by Brown2015). There is evidence for a GC origin for the best-studiedHVSs (e.g. Brown et al. 2012). However, the lack of proper mo-tions or their inaccuracy (Brown et al. 2015) prevents the de-velopment of Galactic trajectories for most HVS stars to tracetheir place of origin. The Hills scenario was challenged by somebrighter HVS B-type stars (e.g. HD 271791, Heber et al. 2008,Przybilla et al. 2008a; HIP 60350, Irrgang et al. 2010) becausethe GC could be excluded as a place of origin. HE 0437–5439is another particularly interesting case, because its time of flightis far too long for it being ejected as a single star from the GC.A possible origin in the Large Magellanic Cloud is under de-bate (Edelmann et al. 2005; Przybilla et al. 2008b; Brown et al.2010, 2015; Irrgang et al. 2013). Perets et al. (2009) suggested

Article published by EDP Sciences A58, page 1 of 12

A&A 601, A58 (2017)

that a close hypervelocity binary could be ejected from a hierar-chical triple through interactions with the SMBH in the GC. Dur-ing their stellar evolution it is possible for such close binaries toevolve to mass transfer configurations, and they may even mergeto form a blue straggler, which would be sufficiently long-lived.

Another mechanism to accelerate stars is the dynamical ejec-tion from open clusters (Leonard 1991). During a close en-counter large kicks can be transferred to the least massive of theinvolved components. This process is most efficient when twoclose binaries collide. Several hundred km s−1 can be reached,but only at rates that cannot account for a significant frac-tion of the observed population of HVSs in the Galactic halo(Perets & Šubr 2012).

Blaauw (1961) first proposed the binary supernova ejectionmechanism. When a massive primary undergoes a core-collapsesupernova explosion, its secondary is released with an ejectionvelocity that is closely connected to the secondary’s orbital ve-locity (Tauris & Takens 1998).

Abadi et al. (2009) predicted that the disruption of satellitegalaxies may contribute halo stars by stripping them from theirhost. The stars may reach velocities exceeding the escape veloc-ity of the Galaxy. This scenario would form a cluster of HVSsin the sky. A large portion of the HVSs from the survey ofBrown et al. (2014) indeed cluster around the constellations ofLeo and Sextans.

2. High-velocity hot subdwarfs

HVSs were also found among hot subdwarf stars. Subluminousstars of spectral type B and O (sdB, sdO) are likely formed outof a red giant star (RG) that has lost almost its entire hydro-gen envelope. The remaining layer of hydrogen does not haveenough mass to sustain a hydrogen-burning shell, like in coolerhorizontal branch stars, and sdO/Bs cannot evolve in the canon-ical way by ascending the asymptotic giant branch before theyfinally settle on the white dwarf cooling tracks (see Heber 2009,2016, for reviews). How the stars are originally stripped of theirhydrogen envelope remains under debate. Systematic surveys re-vealed that a large portion (40–70%) of hot subdwarfs are mem-bers of close binaries (Maxted et al. 2001; Morales-Rueda et al.2003; Copperwheat et al. 2011; Geier et al. 2015b), with mostlywhite dwarfs or low-mass late-type main-sequence stars as com-panions. Substellar companions are also known, however, likebrown dwarfs (Schaffenroth et al. 2015). Wide binaries with F,G, K companions and orbital periods of ∼1000 d exist and maybe formed by stable Roche-lobe overflow (Vos et al. 2012, 2013;Barlow et al. 2013). While the close binaries can be explainedby a common envelope and spiral-in phase during the RG phase,single hot subdwarfs are less straightforward to explain throughmergers of helium white dwarfs, common-envelope mergers, orinternal mixing processes (see Heber 2016). These scenariosare of particular interest to explain the properties of extremelyhelium-rich O-type subdwarfs (He-sdO).

The only known unbound subluminous HVS, US 708, issuch a He-sdO. It was discovered by Hirsch et al. (2005) asthe second HVS. The spectroscopic reobservation of US 708and ground-based proper motion measurements showed that itis the fastest unbound star known so far (Geier et al. 2015a).Ground-based as well as Hubble Space Telescope proper motionmeasurements by Geier et al. (2015a) and Brown et al. (2015)exclude an origin in the GC and therefore the Hills ejectionmechanism. Since the Hills scenario is not valid, a binary su-pernova scenario has been proposed for the ejection of US 708.Geier et al. (2015a) suggested that US 708 is most likely the

ejected donor remnant of a thermonuclear supernova (SN Ia) af-ter it was spun up by the tidal interaction with its former closewhite dwarf companion. Because the orbit shrinks as a result ofthe radiation of gravitational waves, US 708 started to transferhelium-rich matter to its compact white dwarf companion. Aftera critical mass was deposited on the surface of the white dwarf,the helium ignited and triggered the explosion of the C/O coreof the white dwarf. This so-called double detonation has beenproposed as the cause for underluminous SN Ia (Smith et al.2009; Ganeshalingam et al. 2011). Geier et al. (2013) identifiedthe sdB binary CD–30◦ 11223 as a progenitor candidate for sucha scenario. The sdB in CD–30◦ 11223 has been spun up by thetidal influence of the close white dwarf companion to a pro-jected rotational velocity 3rot sin i ' 180 km s−1, which is sig-nificantly higher than the rotation that was found for single sdBs(<10 km s−1, Geier & Heber 2012). An ejected remnant is pre-dicted to have similarly high 3rot sin i (Pan et al. 2013).

The Hyper-MUCHFUSS project was started with the aim tofind potentially unbound hot subdwarfs. Twelve sdB stars havebeen found during the first campaign (Tillich et al. 2011). Oneof the goals is to distinguish between an old bound populationof hot subdwarfs in the Galactic halo and the possibly unboundejected SN Ia donor remnants similar to US 708. In the lattercase, they are predicted to be fast rotators spun up by the tidal in-fluence of their close companions. In the former case, they wereformed as single stars and are expected to be slow rotators justlike the single sdBs in the field (Geier & Heber 2012).

The possibly unbound Hyper-MUCHFUSS sdB SDSSJ121150.27+143716.2 (short J1211) is of particular interestbecause we discovered a cool companion to this sdB star(Németh et al. 2016) orbiting through the outermost parts of theMilky Way. This immediately excludes the SN channel. An ori-gin from the GC and the acceleration there through the sling-shot mechanism was also excluded. First, because the binaryis too wide to have survived the destruction of a hierarchicaltriple. Second, its kinematics in the past do not point to the GC.Németh et al. (2016) suggested the formation in the halo or theaccretion from the tidal debris of a dwarf galaxy that was dis-rupted by the Milky Way (Abadi et al. 2009).

In the following section we present the spectroscopicand kinematic analysis of four interesting hot subdwarfs.The sdB star SDSS J123137.56+074621.7 (J1231) andthe He-sdO star SDSS J205030.39−061957.8 (J2050) havebeen discovered as new objects with extreme kinematics(Geier et al. 2015b). SDSS J163213.05+205124.0 (J1632) andSDSS J164419.44+452326.7 (J1644) have previously been in-vestigated by Tillich et al. (2011) and are now revisited here.They were reobserved with higher quality data to improve theconstraints on their origins and kinematics.

3. Observations, atmospheric parameters,and spectroscopic distances

Preliminary atmospheric parameters, spectroscopic distances,and radial velocities have been obtained from low-resolutionSDSS spectra for the preselection of interesting candidates.For the more accurate analyses presented here we used spec-tra taken with the SDSS/BOSS, Keck/ESI, ESO-VLT/X-shooter,and ESO-VLT/FORS1 spectrographs. The ESI spectra have beenreduced with the pipeline Makee1. Pipeline-reduced BOSS andX-shooter spectra have been downloaded from the SDSS and

1 http://www.astro.caltech.edu/~tb/makee/

A58, page 2 of 12

E. Ziegerer et al.: Fast sdB stars from the Hyper-MUCHFUSS project

Fig. 1. Fit of a model spectrum (full line) of the Balmer series for J1231with the X-shooter observation spectrum (grey).

the ESO Phase 3 databases, respectively. The reduction of theFORS1 spectra is described in Tillich et al. (2011). Details aboutwavelength coverage and resolution of the spectra are providedin Table A.1.

All spectra were used to search for radial velocity variationsin order to search for possible companions. Therefore, we fitted aset of mathematical functions (Gaussians, Lorentzians and poly-nomials) to the hydrogen Balmer lines, and if present, to heliumlines. The FITSB2 routine by Napiwotzki et al. (2004) was ap-plied as well as the spectrum plotting and analysis suite (SPAS)developed by Hirsch (2009). No radial velocity variations weredetected within the uncertainties.

3.1. Atmospheric parameters

A quantitative spectral analysis also provided the atmosphericparameters effective temperature Teff, surface gravity log g,and helium abundance, as well as limits on the projected ro-tational velocity 3rot sin i. We applied the method describedin Lisker et al. (2005) and Stroeer et al. (2007). To determinethe atmospheric parameters, we fitted the Balmer, He i, andHe ii lines with model spectra by means of χ2-minimizationusing the SPAS routine (Hirsch 2009). For the sdB stars withtemperatures Teff lower than 30 000 K (J1231 and J1632) weused a grid of metal line-blanketed local thermal equilibrium(LTE) model spectra of Heber et al. (2000) with solar metallic-ity. For the one star with Teff greater than 30 000 K (J1644) weused LTE model spectra with enhanced metal line-blanketing ofO’Toole & Heber (2006). For the He-sdO star (J2050) we ap-plied the NLTE model spectra of Hirsch & Heber (2009) thattake into account the line-blanketing caused by nitrogen andcarbon. The adopted uncertainties are typical systematic devi-ations between different models (see Geier et al. 2007, for de-tails). The statistical uncertainties based on a bootstrapping al-gorithm are smaller in all cases. The results are listed in Table 1.Figures 1 and 2 show the best fit of a model spectrum with theX-shooter spectrum of J1231 for the region of the Balmer se-ries and HeI lines, respectively. As illustrative examples, Figs. 3and 4 show the best fit of a model spectrum for the region of HeIand HeII lines with the FORS1 and ESI spectrum of J2050.

Fig. 2. Fit of a model spectrum (full line) of HeI lines for J1231 withthe X-shooter observation spectrum (grey).

Fig. 3. Fit of a model spectrum (full line) of He lines for J2050 with theFORS1 observation spectrum (grey).

The three sdB stars have typical effective temperatures. How-ever, it is worth mentioning that the low gravity of J1231 impliesthat the star is close to termination of core helium burning, pos-sibly even beyond that phase. The helium content of J1231 andJ1632 is typical for the majority of sdB stars. However, we couldnot detect any helium lines in the hot J1644, which implies thatits abundance (He/H < 1/1000) is considerably lower than ex-pected for sdBs of similar temperature (Edelmann et al. 2003).The sdO star J2050 does not show any hydrogen, and we wereonly able to derive a lower limit of the helium-to-hydrogen ra-tio of 100. Its temperature, gravity, and helium content are typi-cal for He-sdO stars (Stroeer et al. 2007), in particular similar tothat of the hyper-velocity sdO star US 708 (Geier et al. 2015a).For US 708 an unexpected high projected rotational velocity of3rot sin i = 115 ± 8 km s−1 was found. In comparison to US 708,all program stars show moderate 3rot sin i < 45 km s−1.

A58, page 3 of 12

A&A 601, A58 (2017)

Table 1. Atmospheric parameters.

Name Short Type V AV Teff log g log n(He)log n(H) 3rot sin i 3rad d

(mag) (mag) (K) (cgs) (km s−1) (km s−1) (kpc)

SDSS J123137.56+074621.7 J1231 sdB 17.44 0.05 25 200 ± 500 5.13 ± 0.05 −2.23 ± 0.05 <45 467 ± 2 6.3+0.5−0.5

SDSS J163213.05+205124.0 J1632 sdB 17.62 0.15 28 900 ± 500 5.61 ± 0.05 −1.83 ± 0.03 <33 −239 ± 4 4.3+0.3−0.3

SDSS J164419.44+452326.7 J1644 sdB 17.39 0.03 33 600 ± 500 5.73 ± 0.05 <−3.0 <38 −309 ± 9 4.1+0.3−0.3

SDSS J205030.39–061957.8 J2050 He-sdO 18.22 0.20 47 500 ± 1000 5.70 ± 0.1 >+2.0 <38 −509 ± 19 7.0+0.9−0.8

Notes. V is the apparent magnitude, AV is the reddening in V , and d is the heliocentric distance.

Fig. 4. Fit of a model spectrum (full line) of He lines for J2050 with theESI observation spectrum (grey).

3.2. Spectroscopic distances

From the atmospheric parameters and the apparent visual mag-nitude we derived the spectroscopic distance as described inRamspeck et al. (2001). For J1231 and J1632 we adopted theatmospheric parameters that were obtained from the X-shooterspectra, as they have the highest resolution and a wide wave-length range so that the Balmer jump is accessible, which is verysensitive to the gravity log g. For the remaining two stars weadopted a mean value of the results from ESI and BOSS (J1644),and ESI and SDSS (J2050) spectra, respectively.

The SDSS g and r magnitudes were converted into JohnsonV magnitudes2, which then were corrected for interstellar red-dening. The reddening was found using a dust extinction toolthat gives the Galactic dust reddening for a line of sight3.

3.3. Spectral energy distribution

To check whether the spectroscopic values are consistent withphotometry, we performed a fit of the observed spectral energydistribution (SED). Synthetic SEDs are based on the Atlas12code (Kurucz 1996) using an averaged metal abundance fromFig. 6 in Naslim et al. (2013) as baseline metallicity. While the

2 http://www.sdss.org/dr6/algorithms/sdssUBVRITransform.html3 irsa.ipac.caltech.edu/applications/DUST

effective temperature and surface gravity are fixed to their spec-troscopic values, we fitted the angular diameter as a distancescaling factor, the color excess E(B − V) as a measure of in-terstellar extinction (using the description of Fitzpatrick 1999),and the scaled average abundance pattern. The observed SEDsof our four stars were perfectly matched by the synthetic spectracalculated using the atmospheric parameters derived from spec-troscopy. The obtained values for distances and reddening fromSED-fitting fit to those obtained by spectroscopy within their un-certainties.

3.4. Search for signatures of potential cool companions

Photometric magnitudes from GALEX DR64 and SDSS DR12were available for all stars. BATC DR15, UKIDSS DR9(Lawrence et al. 2007), and ALLWISE (Cutri & et al. 2013)were only available for J1231. Therefore, data in the infraredwere only available for one star, and it was possible only forthis one star to search for an infrared excess as an indicationfor a cooler companion (Fig. 5). There is no sign of a coolercompanion as was seen in the SED of the fast sdB star J1211(Németh et al. 2016). J1211 was analysed in the same way as forour sample, and we found a K-type companion that produced aninfrared excess in the SED. Absorption lines of the companionof J1211 were also visible in the spectrum. No such lines werefound in any of our four program stars. Figure 6 shows a compar-ison of the spectra (X-shooter for J1231, ESI for J1632, J1644,and J2050) of the program stars with the spectrum of J1211.While the spectrum of J1211 shows the Mg i triplet in the re-spective area, none is visible in any of the program stars. Hence,there is no evidence for a cool companion to any of our programstars.

4. Proper motions

The proper motions of the program stars were either takenfrom Tillich et al. (2011) or determined by the same methodas described there. Early-epoch photographic plates from theDigitised Sky Surveys6 were combined with those obtained fromthe data bases of modern digital surveys such as SDSS7, SuperCosmos8, and VHS9. This provided a time base of about 60 yr.

4 Available in the MAST archive: http://galex.stsci.edu/GR6/?page=mastform5 http://vizier.cfa.harvard.edu/viz-bin/Cat?II/2626 http://archive.stsci.edu/cgi-bin/dss_plate_finder7 http://skyserver.sdss3.org/public/en/tools/chart/navi.aspx8 http://www-wfau.roe.ac.uk/sss/pixel.html9 http://www.eso.org/qi/catalogQuery/index/51

A58, page 4 of 12

E. Ziegerer et al.: Fast sdB stars from the Hyper-MUCHFUSS project

8

7

6

5

4

3

2

1

500002000050002000 100001000

0.1

0

-0.1

fλ3

(10−

5 erg

cm−2

s−1

Å2 )

λ (Å)

mx,

mod

el−

mx

(mag

)

W1KH

JY

po

nm

kjig

e

dc

b zi

r

g

uNUV

FUV

Fig. 5. Comparison of synthetic and observed photometry of J1231: the top panel shows the spectral energy distribution. The colored data pointsare fluxes that are converted from observed magnitudes, and the solid grey line is the model. The residual panel at the bottom shows the differencesbetween synthetic and observed magnitudes. The photometric systems have the following color code: GALEX (violet), BATC (gold), SDSS(goldenred), UKIDSS (pink), and WISE (magenta).

Fig. 6. Comparison of the spectra of the four program stars (X-shooterfor J1231, ESI for J1632, J1644, and J2050) with the spectrum of J1211in the area of Mg I triplet (marked with dotted vertical lines).

For each star, positions were derived from all available im-ages with respect to a set of faint, compact, and well-distributedbackground galaxies. The galaxies for the reference system aretaken from the SDSS database. We used as many galaxies as pos-sible, but excluded those that show displacements which couldbe true motion (if the object is misclassified in the SDSS and isin fact a foreground star). The object was then excluded in allepochs. It can also be spurious if it is detected only in certainimages, which can be caused for instance by a close faint back-ground star that is detected only in certain wavelength ranges,

as the photographic plates are taken in different filters and thecompanion is only detected in certain filters. Then these objectsare only excluded for those epochs where the motion occurs.

The comparison of our proper motions with catalogues suchas APOP (Qi et al. 2015), HSOY (Altmann et al. 2017), PP-MXL (Roeser et al. 2010), SDSS (Ahn et al. 2012), USNO-B1.0(Monet et al. 2003), and UCAC4 (Zacharias et al. 2012) showedthat our values are in good agreement within the uncertainties,see Table 2. The HSOY catalogue is a combination of Gaia DR1and PPMXL data. The resulting values are in good agreementwith the values of our proper motion with smaller uncertaintiesthan for PPMXL alone.

For J1644 alone, the values for one of the two proper mo-tion (µα cos δ) components differ between the different measure-ments. Therefore, we discuss two different options for J1644.First, we use the proper motion obtained by Tillich et al. (2011),and second, a weighted mean of the catalogue values (denotedas J1644b). For the remaining program stars we used our propermotion or the one obtained from Tillich et al. (2011) for the fur-ther analysis.

5. Kinematics: extreme halo or ejected stars

We calculate trajectories of the program stars in three differ-ent Milky Way mass models of Irrgang et al. (2013) to trace theorbits back to the Galactic disk to obtain their dynamical prop-erties and possible origins. The halo mass of these three mod-els ranges from MR<200 kpc = 1.2−3.0 × 1012 M�, which cov-ers the whole range of halo masses of other widely used halomass distributions. Nevertheless, we tested a fourth mass model(Rossi et al. 2017). All mass models share the same disk struc-ture (Miyamoto & Nagai 1975). While Irrgang et al. (2013) alsoused their bulge model, Rossi et al. (2017) used the Hernquist(1990) model. Model III of Irrgang et al. (2013) and Rossi et al.(2017) used the same potential form for the halo, namely theone suggested by Navarro et al. (1997). However, we recall thatthe mass model of Rossi et al. (2017) was calibrated to different

A58, page 5 of 12

A&A 601, A58 (2017)

Table 2. Proper motions.

Name µα cos δ µδ Catalogue( mas yr−1) ( mas yr−1)

J1231 −7.9 ± 3.4 −5.0 ± 2.8 this paper−2.5 ± 1.3 −6.8 ± 1.9 APOP−10.5 ± 2.5 −7.0 ± 2.5 HSOY−7.5 ± 5.6 −4.1 ± 5.6 PPMXL−4 ± 3 −2 ± 3 SDSS

J1632 −12.5 ± 3.0 −1.6 ± 3.6 T11−8 ± 2.7 −3.5 ± 3.4 APOP−16.6 ± 5.3 −5.8 ± 5.3 PPMXL−13 ± 3 −4 ± 3 SDSS−10 ± 2 0 ± 2 USNO-B1.0

J1644 4.7 ± 2.8 −26.1 ± 3.3 T11−1.1 ± 3.2 −16.4 ± 3.2 APOP−7 ± 5.9 −27.5 ± 5.9 PPMXL−1 ± 3 −26 ± 3 SDSS−2 ± 6 −26 ± 3 USNO-B1.0

J2050 5.5 ± 4.8 −8.9 ± 3.5 this paper3.2 ± 3.7 −2.9 ± 2.1 APOP1.8 ± 2.4 −9.7 ± 2.4 HSOY−3.8 ± 6.2 −7.5 ± 6.2 PPMXL

0 ± 3 −4 ± 3 SDSS

Notes. T11: Tillich et al. (2011).

observational constraints than the mass models of Irrgang et al.(2013), which leads to different halo masses of the two massmodels.

Long-term orbits were calculated for 5000 Myr to charac-terise them in the context of population synthesis. In order toconstrain the place of origin, that is, to determine whether thestar was ejected from the Galactic disk or centre, we tracedthe trajectories back to their last disk crossings and calculatedthe times of flight and ejection velocities for all mass models.Through a Monte Carlo simulation of a Gaussian distributionwith a depth of 106, we determined all kinematic parameters ofthe current location of the stars as well as the values at the timeand position of their last disk passage, such as velocity com-ponents in Cartesian coordinates (3x, 3y, 3z), with the Sun ly-ing on the negative x-axis and the north Galactic pole being onthe positive z-axis. Cylindrical coordinates (3r, 3φ, 3z), Galacticrest-frame velocity 3grf, and ejection velocity 3ej corrected forthe Galactic rotation were also calculated for each of the fourmass models. The input parameters for the simulation are theradial velocity 3rad, proper motions (µα cos δ and µδ), and spec-troscopic distance d with their corresponding uncertainties. Forall program stars the resulting disk passage is independent of thechoice of the applied mass model.

From the long-term calculations the z-component of the an-gular momentum Jz as well as the eccentricity e of the orbit aredetermined. All resulting velocity components and the probabil-ity of being bound for models I, II, and III of Irrgang et al. (2013)and the model of Rossi et al. (2017) can be found in Table 3 (1σuncertainties are given). As can be seen, the choice of model po-tential is of no importance because all velocities derived fromthe different models agree within their mutual uncertainties.While J1632 is certainly bound to the Galaxy, the probability

J1644b

J1231J2050

J1644

J1632

3002001000-100-200-300-400

300

200

100

0

-100

-200

-300

-400

-500

3φ (km s−1)

3 r(k

ms−

1 )

Fig. 7. Comparison of the program stars in the 3r − 3φ-diagram with 3σcontour of the thick disk (dashed line) and 3σ contour of the thin disk(solid line) according to Pauli et al. (2006).

that J1231, J1644, and J2050 are unbound is also low, regard-less of the choice of Galactic potential. Therefore we concludethat our program stars belong to an old Galactic stellar popula-tion and investigate their kinematical properties from long-termevolution of their Galactic orbits.

According to Pauli et al. (2006), stars can be assigned to thepopulations of the different components of the Milky Way – thindisk, thick disk, halo – using three different criteria. The first isthe classification by their position in the 3r − 3φ-diagram (Fig. 7),where 3r is the Galactic radial component, which is negative to-wards the GC, while 3φ is the Galactic rotational component.Stars that are revolving on retrograde orbits around the GC havenegative 3φ. Disk stars are located in a well-defined region. Thinand thick disk overlap. Stars that are outside this region are as-sumed to belong to the Galactic halo. Figure 7 shows the posi-tion of the program stars in the 3r − 3φ-diagram compared to 3σcontours of the thick and thin disk as introduced by Pauli et al.(2006). All stars lie well outside the disk region and can there-fore be considered as halo stars.

The second diagnostic tool is the Jz − e-diagram, which isshown in Fig. 8. Stars on retrograde orbits have positive Jz. Thin-disk stars are located at the top left end of the diagram, havingvery low eccentricities e. Tillich et al. (2011) suggested that starsinside the box belong to the thick disk, while stars inside theellipse are typical halo stars as they show only little effect of thedisk rotation and cross the Galactic plane almost perpendicularon highly eccentric orbits. Again, our stars lie well outside thedisk region.

The third classification criterion is the shape of the orbit inthe r − z-diagram itself, where r is the distance of the star to theGC projected onto the Galactic plane r =

√x2 + y2. Thin-disk

orbits only cover a very narrow region in this diagram becausethey are on very low-eccentricity orbits with very low inclina-tion. They vary in r by less than 3 kpc and in z by less than1–2 kpc. Thick-disk stars show a larger spread in both variables.Halo objects can have any chaotic orbit imaginable. The orbitsof our stars are discussed individually in the following sections.For each star the average orbit from the Monte Carlo simulationwas calculated 5000 Myr into the past.

A58, page 6 of 12

E. Ziegerer et al.: Fast sdB stars from the Hyper-MUCHFUSS project

Table 3. Velocity components.

Name/ 3x 3y 3z 3r 3φ 3GRF 3ej BoundModel (km s−1) (km s−1) (km s−1) (km s−1) (km s−1) (km s−1) (km s−1)J1231 −66 ± 97 −136 ± 86 378 ± 29 99 ± 102 −114 ± 81 428 ± 32I −157 ± 105 −144 ± 76 429 ± 27 128 ± 102 −152 ± 111 495 ± 51 611 ± 67 99.9%II −156 ± 105 −146 ± 76 429 ± 27 127 ± 102 −154 ± 111 495 ± 51 612 ± 67 99.3%III −154 ± 104 −146 ± 76 429 ± 27 125 ± 102 −154 ± 110 494 ± 51 612 ± 67 100%R −153 ± 106 −147 ± 76 435 ± 27 124 ± 102 −155 ± 111 499 ± 51 617 ± 67 99.5%J1632 −179 ± 58 −39 ± 61 33 ± 50 155 ± 67 −97 ± 51 203 ± 54I −305 ± 58 144 ± 79 256 ± 63 −4 ± 143 −304 ± 103 435 ± 62 612 ± 65 100%II −306 ± 60 152 ± 82 254 ± 64 −11 ± 145 −309 ± 104 438 ± 63 616 ± 65 100%III −302 ± 58 131 ± 76 263 ± 64 16 ± 134 −300 ± 96 432 ± 65 609 ± 65 100%R −317 ± 70 185 ± 113 281 ± 90 39 ± 153 −339 ± 114 481 ± 92 661 ± 90 100%J1644 433 ± 72 −83 ± 41 −257 ± 42 −432 ± 73 84 ± 41 514 ± 69I −467 ± 117 −66 ± 104 430 ± 82 40 ± 252 402 ± 145 660 ± 51 553 ± 50 91.5%II −451 ± 120 −72 ± 112 430 ± 88 72 ± 250 381 ± 149 652 ± 52 551 ± 53 79.3%III −489 ± 118 −40 ± 95 435 ± 78 −1 ± 267 410 ± 155 675 ± 58 568 ± 55 100%R −489 ± 90 −48 ± 128 372 ± 102 338 ± 181 304 ± 162 640 ± 64 572 ± 44 84.5%J1644b 355 ± 70 −142 ± 51 −162 ± 55 −382 ± 72 0 ± 49 422 ± 69I −408 ± 148 116 ± 124 247 ± 159 312 ± 252 97 ± 218 527 ± 161 546 ± 116 99.5%II −396 ± 153 114 ± 125 237 ± 159 316 ± 239 87 ± 209 511 ± 168 537 ± 120 97.9%III −424 ± 136 132 ± 119 267 ± 159 302 ± 277 90 ± 231 551 ± 149 570 ± 112 100%R −488 ± 90 −48 ± 128 372 ± 102 338 ± 181 304 ± 162 640 ± 64 572 ± 44 84.5%J2050 −299 ± 73 −191 ± 71 107 ± 92 52 ± 139 −332 ± 66 394 ± 40I −114 ± 149 46 ± 74 −104 ± 35 −94 ± 144 −107 ± 41 215 ± 92 385 ± 79 99.8%II −120 ± 146 41 ± 74 −102 ± 35 −101 ± 142 −104 ± 41 215 ± 93 383 ± 80 99.6%III −91 ± 160 53 ± 79 −107 ± 38 −69 ± 157 −110 ± 39 217 ± 92 389 ± 78 100%R −140 ± 140 33 ± 76 −112 ± 35 −118 ± 138 −106 ± 43 226 ± 94 390 ± 82 99.7%

Notes. Velocity components and the probability of being bound to the Galaxy of the program stars. The values of the first line are the currentvalues, next lines are the values at the last disk passage based on models I, II, and III of Irrgang et al. (2013) and the model of Rossi et al. (2017,R), respectively.

J1644b

J1231

J2050

J1632

J1644

10.80.60.40.204000

3000

2000

1000

0

-1000

-2000

-3000

e

J z(k

pckm

s−1 )

Fig. 8. Jz−e-diagram, the dashed line indicates the region of typical halostars. The solid line marks the thick-disk region. Thin-disk stars wouldpopulate the continuation of the parallelogram to lower eccentricities.

5.1. J1644 – an extreme halo star

The fastest of the program stars is also the most precarious.Because of the discrepancy in proper motions (see Sect. 4), wecarried out the kinematic analyses twice, adopting the propermotions of Tillich et al. (2011) and a weighted mean of the cat-alogue values, respectively.

Regardless of the choice of the proper motion, J1644 hasan extreme kinematic behaviour, as becomes obvious from its

position in the 3r − 3φ- and the Jz − e-diagrams (see Figs. 7and 8). However, the orbit strongly depends on the choice oftheir values, as demonstrated in the left panel of Fig. 9. Whenwe adopt the proper motion of Tillich et al. (2011), J1644 is ona highly eccentric orbit, which leads the star to distances of upto 129 ± 73 kpc away from the GC. The travel time since its lastapproach to the GC is much longer, 1558±988 Myr, than the life-time of an EHB star. If J1644 were ejected from the GC, it wouldhave been a main-sequence star or subgiant at the time and hadto evolve into an sdB on the way. Adopting the weighted mean ofthe catalogue proper motion values leads to a shorter travel timeof only 113 ± 72 Myr and reaches only distances of 20 ± 6 kpcaway from the GC, which is consistent with the EHB lifetime of<100 Myr, meaning that it is possible to reach the star’s currentposition within the lifetime. The right-hand panel of Fig. 9 showsthat the GC lies within the 1σ contours of the disk passages, re-gardless of the choice of the proper motion. Although J1644 hasthe highest 3grf = 514 ± 69 km s−1 (or 3grf = 422 ± 69 km s−1

for the weighted mean of the catalogue proper motions) of allprogram stars, it is heading towards the GC and must thereforestill be bound.

5.2. Possibly ejected stars

Tillich et al. (2011) suggested that halo stars outside the ellipsein the Jz−e-diagram (Fig. 8) could be ejected stars. Accordingly,

A58, page 7 of 12

A&A 601, A58 (2017)

80

60

40

20

0

-20

-40

-60

-80

20

15

10

5

0

-5

-10

-15

-20

-25

1000

500

200

100

50

20

10

5

21

120100806040200

80

60

40

20

0

-20

-40

-60

-80

20100-10-20

20

15

10

5

0

-5

-10

-15

-20

-25

500

200

100

50

20

10

5

21

z(k

pc)

y(k

pc)

Num

ber

ofm

easu

rem

ents

per

bin

r (kpc)

z(k

pc)

x (kpc)

y(k

pc)

Num

ber

ofm

easu

rem

ents

per

bin

⋆

⊙

⋆

⊙

Fig. 9. Left panel: r−z-diagram of J1644 using the Tillich et al. (2011) proper motion (top) and the weighted mean of the catalogue proper motions(bottom), respectively. Right panel: disk passages binned and colour-coded of J1644 with 1 and 3σ contours using the Tillich et al. (2011) propermotion (top) and the weighted mean of the catalogue proper motions (bottom), respectively. The black dot marks the GC, the star the currentposition of J1644, and the solar symbol the position of the Sun. The circle indicates the Galactic disk. All calculations were performed withmodel I of Irrgang et al. (2013).

J1231, J1632, and J2050, all on retrograde orbits (see Fig. 7),could be runaway stars from the Galactic disk or bound HVSfrom the GC rather than extreme halo stars. While J1231, andJ2050 cannot originate from the GC (see Figs. 11 and 12, rightpanels), the disk-crossing area of the trajectories of J1632 in-clude the GC (see Fig. 10 right panel). We discuss this objectfirst before addressing the disk runaways J1231 and J2050.

5.2.1. J1632 – a potentially bound HVS

The analysis of its trajectory indicates that J1632 may originatefrom the GC and therefore could be a bound HVS (see Fig. 10left panel). J1632 has a relatively low 3GRF = 203 ± 54 km s−1

of the order of typical disk stars and is approaching us. Sim-ilarly, the velocity perpendicular to the Galactic disk is verylow (3z = 33 ± 50 km s−1), similar to that of a thick-disk star.In addition, the eccentricity speaks for a thick-disk star. There-fore the orbit looks like that of a typical thick-disk star (see

left-hand panel of Fig. 10). However, the star is revolving ret-rograde around the GC, and consequently, it cannot be an ordi-nary thick-disk star. Randall et al. (2015) found an intermediateHe-sdB on a similar orbit. An origin from the GC for J1632 isconceivable (see right-hand panel of Fig. 10). Possibly, J1632could have been ejected into a low-inclination orbit when theformer binary was disrupted by the SMBH. With a travel time of23.7 ± 5.4 Myr from the GC to its current position, this scenariois consistent with the EHB lifetime of such stars of <100 Myr.

5.2.2. J1231 and J2050 – potential disk runaways

The constant radial velocity, proper motion, and spectroscopicdistance of J1231 indicate a Galactic rest-frame velocity of vgrf =

428 ± 32 km s−1 and a likely origin in the Galactic disk ratherthan the GC (see right-hand panel of Fig. 11). A travel timeof 14.4 ± 1.6 Myr from the Galactic disk to its current positionis consistent with the EHB lifetime of such stars of <100 Myr.

A58, page 8 of 12

E. Ziegerer et al.: Fast sdB stars from the Hyper-MUCHFUSS project

151050

10

8

6

4

2

0

-2

-4

-6

-8

-1020100-10-20

20

15

10

5

0

-5

-10

-15

-20

-25

2000

1000

500

200

100

50

20

10

5

21

r (kpc)

z(k

pc)

x (kpc)y

(kpc

)

Num

ber

ofm

easu

rem

ents

per

bin

⋆⊙

Fig. 10. Left panel: r − z-diagram of J1632. Right panel: disk passages binned and colour-coded of J1632 with 1 and 3σ contours. The blackdot marks the GC, the star the current position of J1632, and the solar symbol the position of the Sun. The circle indicates the Galactic disk. Allcalculations were performed with model I of Irrgang et al. (2013).

6050403020100

60

50

40

30

20

10

0

-10

-20

-30

-40

-50

-6020100-10-20

20

15

10

5

0

-5

-10

-15

-20

-25

200

100

50

20

10

521

r (kpc)

z(k

pc)

x (kpc)

y(k

pc)

Num

ber

ofm

easu

rem

ents

per

bin

⋆⊙

Fig. 11. Left panel: r − z-diagram of J1231. Right panel: disk passages binned and colour-coded of J1231 with 1 and 3σ contours. The blackdot marks the GC, the star the current position of J1231, and the solar symbol the position of the Sun. The circle indicates the Galactic disk. Allcalculations were performed with model I of Irrgang et al. (2013).

Although the star is the only one of the program stars recedingfrom us, it is bound with a probability of 99.9%. In the context ofpopulation membership, J1231 shows a quite chaotic orbit likethat of an extreme halo star (see left-hand panel of Fig. 11). Itreaches distances of more than 50 kpc away from the Galacticdisk.

The traced orbits of J2050 show that the star does not ap-proach anywhere near the GC (see right-hand panel of Fig. 12)with a typical halo orbit (see left-hand panel Fig. 12). Withvgrf = 394 ± 40 km s−1 on a retrograde orbit, it has a probabil-ity of being bound of 99.8%. Its travel time from the outskirts ofthe Galactic disk is 113 ± 72 Myr, which is consistent with thelifetime of EHB stars.

6. Conclusions

We have performed a spectroscopic and kinematic follow-up analysis of two known hot subdwarfs from the first

Hyper-MUCHFUSS campaign as well as two new ones with ex-treme kinematics. Radial velocity measurements, spectral identi-fication, and photometry (when available) were used to excludebinarity or variability of the stars. Proper motions were eithertaken from Tillich et al. (2011) or measured in the same way.The goal of this work was to place constraints on the place oforigin of the stars and the possible mechanisms that led the starsto their extreme kinematics.

While we cannot rule out that the program stars could be gen-uine halo stars on extreme Galactic orbits, we considered the rel-evance of three ejection scenarios for our program stars, that is,the Hills scenario, the binary supernova scenario, and a potentialextragalactic origin. The Hills slingshot scenario may be validonly for two of our program stars because their last disk pas-sages came close to the GC (J1632 and J1644). The lifetime ofEHB stars is about 100 Myr. If the stars have been formed in a bi-nary and then have been disrupted by the SMBH, the travel timefrom the GC to their current position must be consistent with

A58, page 9 of 12

A&A 601, A58 (2017)

403020100

20

15

10

5

0

-5

-10

-15

-2020100-10-20

20

15

10

5

0

-5

-10

-15

-20

-25

50

20

10

5

21

r (kpc)

z(k

pc)

x (kpc)

y(k

pc)

Num

ber

ofm

easu

rem

ents

per

bin

⋆

⊙

Fig. 12. Left panel: r − z-diagram of J2050. Right panel: disk passages binned and colour-coded of J2050 with 1 and 3σ contours. The blackdot marks the GC, the star the current position of J2050, and the solar symbol the position of the Sun. The circle indicates the Galactic disk. Allcalculations were performed with model I of Irrgang et al. (2013).

this lifetime. This is the case for J1632. For J1644 this is onlythe case if we adopt the weighted mean of the catalogue propermotions, however. When we adopt the proper motion we mea-sured on our own, the travel time is far too long, which means,if this scenario is true, that the star must have evolved to an sdBafter the former binary was disrupted by the SMBH through oneof the single evolution channels for hot subdwarfs. Another op-tion is the disruption of a hierarchical triple by the SMBH andthe subsequent production of an sdB through the merger of twohelium white dwarfs. Alternatively, the star has evolved to ansdB with a low-mass companion, such as a planet, that probablydid not survive the common-envelope phase. Accurate astrom-etry by Gaia will solve this uncertainty in the proper motionmeasurements. Figure 13 shows how the area of disk passagesshrinks when the uncertainties in proper motion are reduced. Anuncertainty of 0.1 mas yr−1 was also applied, which is a realisticuncertainty that the Gaia mission will provide (de Bruijne 2012).

As the star with the highest velocity known (US 708,Geier et al. 2015a) is a hot subdwarf that was not acceleratedby the slingshot mechanism but rather a supernova explosionin a close binary, this is the scenario that should be considerednext. J2050 is a spectroscopic twin of US 708 and therefore apromising candidate of a surviving secondary of a supernova,as proposed for US 708. It could be originating from a systemsimilar to CD–30◦ 11223 (Geier et al. 2013). For J2050 the ejec-tion velocity 3ej = 385 ± 79 km s−1 and the 3rot sin i < 38 km s−1

are both moderate in comparison to the values of US 708:3ej = 998 ± 68 km s−1, 3rot sin i = 115 ± 8 km s−1. As subd-warfs in compact binaries are assumed to have been spun up bythe tidal influence, the progenitor system does not need to havebeen as tight as CD–30◦ 11223. The progenitor system of J2050could have had properties similar to that of the sdB + WD binaryKPD 1930+2752 (Maxted et al. 2000; Geier et al. 2007). In thissystem, the time in which the two objects will merge as a resultof the radiation of gravitational waves is about twice as long asthe lifetime of the sdB on the EHB. Owing to the shrinkage ofthe orbit, Roche-lobe overflow might be possible before the sdBevolves into a white dwarf (Geier et al. 2007). The travel time

σµ = 2.0 mas yr−1

σµ = 1.5 mas yr−1

σµ = 1.0 mas yr−1

σµ = 0.5 mas yr−1

σµ = 0.1 mas yr−1

50-5-10

8

6

4

2

0

-2

-4

-6

x (kpc)

y(k

pc)

•

⋆

⊙

Fig. 13. Disk passage 3σ contour of J1644 using the Tillich et al. (2011)proper motion as in Fig. 9 (upper right) and for uncertainties σµ reducedfrom 2.0 mas yr−1 to 0.1 mas yr−1. The latter is expected to be the Gaiaend-of-mission accuracy. The black dot marks the GC, the star the cur-rent position of J1644, and the solar symbol the position of the Sun.

from the disk to the current position of J2050 is consistent withthe lifetime. The same is true for the potential disk runaway sdBstar J1231.

The accretion scenario has been proposed by Németh et al.(2016) in order to explain the origin of the binary sdB J1211.According to this, J1211 was accreted from the debris of a de-stroyed satellite galaxy. This scenario could also be valid for ourprogram stars. If this is the case, the stars should belong to stel-lar streams in the halo that are yet to be discovered from Gaiaastrometry.

A58, page 10 of 12

E. Ziegerer et al.: Fast sdB stars from the Hyper-MUCHFUSS project

Acknowledgements. E.Z. acknowledges funding by the German Science foun-dation (DFG) through grant HE1356/45-2. We thank John E. Davis for the de-velopment of the slxfig module that we used to prepare the figures in this pa-per. Some of the data presented in this paper were obtained from the MikulskiArchive for Space Telescopes (MAST). STScI is operated by the Associationof Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. Support for MAST for non-HST data is provided by the NASA Officeof Space Science via grant NNX09AF08G and by other grants and contracts.This work is based in part on data obtained as part of the UKIRT Infrared DeepSky Survey. This publication makes use of data products from the Wide-field In-frared Survey Explorer, which is a joint project of the University of California,Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technol-ogy, funded by the National Aeronautics and Space Administration. Based onobservations collected at the European Organisation for Astronomical Researchin the Southern Hemisphere under ESO program 093.D-0127(A). This work isbased on observations obtained at the W. M. Keck Observatory, which is oper-ated as a scientific partnership among the California Institute of Technology, theUniversity of California, and the National Aeronautics and Space Administra-tion. The Observatory was made possible by the generous financial support ofthe W. M. Keck Foundation. The authors wish to recognize and acknowledgethe very significant cultural role and reverence that the summit of Maunakea hasalways had within the indigenous Hawaiian community. We are most fortunateto have the opportunity to conduct observations from this mountain.

ReferencesAbadi, M. G., Navarro, J. F., & Steinmetz, M. 2009, ApJ, 691, L63Ahn, C. P., Alexandroff, R., Allende Prieto, C., et al. 2012, ApJS, 203, 21Altmann, M., Roeser, S., Demleitner, M., Bastian, U., & Schilbach, E. 2017,

A&A, 600, L4Barlow, B. N., Liss, S. E., Wade, R. A., & Green, E. M. 2013, ApJ, 771, 23Blaauw, A. 1961, Bull. Astron. Inst. Netherlands, 15, 265Brown, W. R. 2015, ARA&A, 53, 15Brown, W. R., Geller, M. J., Kenyon, S. J., & Kurtz, M. J. 2005, ApJ, 622, L33Brown, W. R., Geller, M. J., Kenyon, S. J., Kurtz, M. J., & Bromley, B. C. 2007,

ApJ, 671, 1708Brown, W. R., Anderson, J., Gnedin, O. Y., et al. 2010, ApJ, 719, L23Brown, W. R., Cohen, J. G., Geller, M. J., & Kenyon, S. J. 2012, ApJ, 754, L2Brown, W. R., Geller, M. J., & Kenyon, S. J. 2014, ApJ, 787, 89Brown, W. R., Anderson, J., Gnedin, O. Y., et al. 2015, ApJ, 804, 49Copperwheat, C. M., Morales-Rueda, L., Marsh, T. R., Maxted, P. F. L., & Heber,

U. 2011, MNRAS, 415, 1381Cutri, R. M., & et al. 2013, VizieR Online Data Catalog: II/328de Bruijne, J. H. J. 2012, Ap&SS, 341, 31Edelmann, H., Heber, U., Hagen, H.-J., et al. 2003, A&A, 400, 939Edelmann, H., Napiwotzki, R., Heber, U., Christlieb, N., & Reimers, D. 2005,

ApJ, 634, L181Fitzpatrick, E. L. 1999, PASP, 111, 63Ganeshalingam, M., Li, W., & Filippenko, A. V. 2011, MNRAS, 416, 2607Geier, S., & Heber, U. 2012, A&A, 543, A149Ghez, A. M., Salim, S., Hornstein, S. D., et al. 2005, ApJ, 620, 744Geier, S., Nesslinger, S., Heber, U., et al. 2007, A&A, 464, 299Geier, S., Marsh, T. R., Wang, B., et al. 2013, A&A, 554, A54Geier, S., Fürst, F., Ziegerer, E., et al. 2015a, Science, 347, 1126Geier, S., Kupfer, T., Heber, U., et al. 2015b, A&A, 577, A26Gillessen, S., Eisenhauer, F., Trippe, S., et al. 2009, ApJ, 692, 1075Han, Z. 2008, A&A, 484, L31Heber, U. 2009, ARA&A, 47, 211

Heber, U. 2016, PASP, 128, 082001Heber, U., Reid, I. N., & Werner, K. 2000, A&A, 363, 198Heber, U., Edelmann, H., Napiwotzki, R., Altmann, M., & Scholz, R.-D. 2008,

A&A, 483, L21Hernquist, L. 1990, ApJ, 356, 359Hills, J. G. 1988, Nature, 331, 687Hirsch, H., & Heber, U. 2009, J. Phys. Conf. Ser., 172, 012015Hirsch, H. A. 2009, Ph.D. Thesis, Friedrich-Alexander University Erlangen-

NürnbergHirsch, H. A., Heber, U., O’Toole, S. J., & Bresolin, F. 2005, A&A, 444, L61Irrgang, A., Przybilla, N., Heber, U., Nieva, M. F., & Schuh, S. 2010, ApJ, 711,

138Irrgang, A., Wilcox, B., Tucker, E., & Schiefelbein, L. 2013, A&A, 549, A137Kurucz, R. L. 1996, in Model Atmospheres and Spectrum Synthesis, eds. S. J.

Adelman, F. Kupka, & W. W. Weiss (San Francisco: ASP), 160Lawrence, A., Warren, S. J., Almaini, O., et al. 2007, MNRAS, 379, 1599Leonard, P. J. T. 1991, AJ, 101, 562Lisker, T., Heber, U., Napiwotzki, R., et al. 2005, A&A, 430, 223Madigan, A.-M., Pfuhl, O., Levin, Y., et al. 2014, ApJ, 784, 23Maxted, P. F. L., Marsh, T. R., & North, R. C. 2000, MNRAS, 317, L41Maxted, P. F. L., Heber, U., Marsh, T. R., & North, R. C. 2001, MNRAS, 326,

1391Miyamoto, M., & Nagai, R. 1975, PASJ, 27, 533Monet, D. G., Levine, S. E., Canzian, B., et al. 2003, AJ, 125, 984Morales-Rueda, L., Maxted, P. F. L., Marsh, T. R., North, R. C., & Heber, U.

2003, MNRAS, 338, 752Napiwotzki, R., Karl, C. A., Lisker, T., et al. 2004, Ap&SS, 291, 321Naslim, N., Jeffery, C. S., Hibbert, A., & Behara, N. T. 2013, MNRAS, 434,

1920Navarro, J. F., Frenk, C. S., & White, S. D. M. 1997, ApJ, 490, 493Németh, P., Ziegerer, E., Irrgang, A., et al. 2016, ApJ, 821, L13O’Toole, S. J., & Heber, U. 2006, A&A, 452, 579Pan, K.-C., Ricker, P. M., & Taam, R. E. 2013, ApJ, 773, 49Pauli, E.-M., Napiwotzki, R., Heber, U., Altmann, M., & Odenkirchen, M. 2006,

A&A, 447, 173Perets, H. B., & Šubr, L. 2012, ApJ, 751, 133Perets, H. B., Wu, X., Zhao, H. S., et al. 2009, ApJ, 697, 2096Przybilla, N., Fernanda Nieva, M., Heber, U., & Butler, K. 2008a, ApJ, 684,

L103Przybilla, N., Nieva, M. F., Heber, U., et al. 2008b, A&A, 480, L37Qi, Z., Yu, Y., Bucciarelli, B., et al. 2015, AJ, 150, 137Ramspeck, M., Heber, U., & Moehler, S. 2001, A&A, 378, 907Randall, S. K., Bagnulo, S., Ziegerer, E., Geier, S., & Fontaine, G. 2015, A&A,

576, A65Roeser, S., Demleitner, M., & Schilbach, E. 2010, AJ, 139, 2440Rossi, E. M., Marchetti, T., Cacciato, M., Kuiack, M., & Sari, R. 2017, MNRAS,

467, 1844Schaffenroth, V., Barlow, B. N., Drechsel, H., & Dunlap, B. H. 2015, A&A, 576,

A123Schödel, R., Ott, T., Genzel, R., et al. 2003, ApJ, 596, 1015Smith, G. P., Ebeling, H., Limousin, M., et al. 2009, ApJ, 707, L163Stroeer, A., Heber, U., Lisker, T., et al. 2007, A&A, 462, 269Svensson, K. M., Church, R. P., & Davies, M. B. 2008, MNRAS, 383, L15Tauris, T. M., & Takens, R. J. 1998, A&A, 330, 1047Tillich, A., Heber, U., Geier, S., et al. 2011, A&A, 527, A137Vos, J., Østensen, R. H., Degroote, P., et al. 2012, A&A, 548, A6Vos, J., Østensen, R. H., Németh, P., et al. 2013, A&A, 559, A54Zacharias, N., Finch, C. T., Girard, T. M., et al. 2012, VizieR Online Data

Catalog: I/322

A58, page 11 of 12

A&A 601, A58 (2017)

Appendix A: Additional table

Table A.1. Atmospheric parameters, radial velocities, and projected rotational velocities.

Name OBS Teff log g log n(He)n(H) 3rot sin i 3rad

(K) (cgs) (km s−1) (km s−1)

J1231 SDSS-BOSS 25 200 ± 400 5.11 ± 0.04 −2.29 ± 0.16 460 ± 8

X-shooter/VLTb 25 200 ± 500 5.13 ± 0.02 −2.23 ± 0.05 <45 467 ± 2

J1632 SDSSa 26 870 ± 610 5.31 ± 0.09 −2.1 ± 0.2 −239 ± 10

SDSS-BOSS 29 000 ± 500 5.46 ± 0.06 −1.59 ± 0.09 −261 ± 20

ESI/Keck 29 500 ± 400 5.61 ± 0.07 −1.78 ± 0.04 <35 −253 ± 10

X-shooter/VLTb 28 900 ± 500 5.61 ± 0.02 −1.83 ± 0.03 <33 −239 ± 2

J1644 SDSSa 31 680 ± 410 5.78 ± 0.11 −2.9 ± 0.3 −314 ± 5

SDSS-BOSSb 33 400 ± 200 5.69 ± 0.04 <3.0 −309 ± 9

ESI/Keckb 33 800 ± 200 5.76 ± 0.04 <−3.0 <38 −299 ± 10

J2050 SDSSb 48 000 ± 500 5.68 ± 0.05 >+1.3 −509 ± 19

FORS1/VLT 48 600 ± 700 5.84 ± 0.12 >+2.0 −485 ± 44

ESI/Keckb 47 000 ± 200 5.71 ± 0.06 >+2.0 <38 −473 ± 10

Notes. SDSS-BOSS: R = 2200, 3600–10 000 Å, ESI: echellette mode with 0.5 arcsec-slit, R = 8000, 4000–6000 Å, X-shooter: R = 10 000,3000–6800 Å, FORS1: R = 1800, 3730–5200 Å. (a) Values are taken from Tillich et al. (2011). (b) Values adopted for the kinematic calculations inthis paper.

A58, page 12 of 12